arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2512.08149math.LOcs.CCcs.LO
收藏

均匀类分离中的对抗障碍

Adversarial Problems and the Limits of Uniform Resolution

Milan Rosko

AI总结:

本文研究了构造算术中均匀类分离的结构障碍,指出在特定条件下,均匀分离原则成为固定点构造的特例,并限制了HA内分离的逻辑形式。

摘要:

我们识别出一个强大的结构障碍,阻碍了构造算术中的均匀分离。该机制与语义内容无关;当两个不同的评估谓词并行存在且推理仍能在HA的扩展中均匀表示时,该障碍会出现。在这些条件下,任何可能的均匀类分离原则都成为固定点构造的一个特例。所导致的限制在范围上比经典分离障碍(Baker;Rudich;Aaronson等)更严格,因为它限制了HA内的均匀分离的逻辑形式,而不是限制特定的相对化、自然化或代数化技术。

英文摘要:

A general method of resolution can take part in forming problems once their language represents the method itself. A fixed formation rule gives each sound resolver a presentation it cannot answer. Admitting the diagnosis yields a sound successor with an instance of its own. For standard arithmetical theories $\mathrm{T}$, the instance is $\mathsf{HA}$-provably equivalent to $\mathrm{Con}(\mathrm{T})$. A theory that decides its own instance asserts a false sentence. The $m$-th iterated independence question is equivalent to the instance $m$ stages later. No effective admission policy with consistent union closes the regress. A noncomputable answer function resolves all program indices, while a false assertion of resolution supplies no warrant. At feasible scale, bounded self-consistency is cheap. A clocked Gödel sentence gives an unconditional gap relative to its short binary description, but no superpolynomial lower bound in the window size. The arithmetic identity motivates studying bounded consistency of the successor theory, as in a conjecture of Pudlák. We state its finitely axiomatized case and prove that it implies $\mathcal{NP}\ne\mathrm{co}\mathcal{NP}$. The conjecture motivates a feasible analogue, but does not establish all clauses of the unbounded barrier.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑